教学课程资源库 更多>>

关键词： Morphology   Design   Animal sciences   Computer science   Entomology   Fluid mechanics   Industrial engineering   Mechanics   Optics   Physical geography   Physics   Polymer chemistry  

摘要： Bioinspired Micro/Nano textured rough topography of a surface has many applications in super-hydrophobicity, self-cleaning surface, anti-icing coatings, anti-biofouling, and drag reduction surfaces. The role of hierarchical and complex surface topography in nature is to amplify the hydrophobicity and maximize the fouling resistance. While there are various manufacture methods to produce these rough topography surfaces, there are still overarching challenges including the cost and manufacturing scalability attributed to the photolithography-based microfabrication processes. Thus, a similar micro- and nano-scale 3-D topographic surfaces inspired from the nature were fabricated using a simple and scalable process. In this study, we utilized an inexpensive and scalable sheet-to-sheet roll coating method to produce a durable surface based on the carbon nanotube reinforced polymer nanocomposite. The surface textures consisting of micro and nano-scale hierarchical structures were produced due to the transverse instabilities of the non-Newtonian viscoelastic polymer fluid called ribbing. This ribbing instability produces a three-dimensional wavy deformation during the shearing between the rollers. The polymer composite retains the deformed shape due to the recovery of highviscosity after removing the shear stress. The nanocomposite material used in this study was synthesized by dispersing 10 wt % of multiwalled carbon nanotube (MWCNT, 10-nm diameter, and 100-μm length) in polydimethylsiloxane (PDMS) polymer. This non-Newtonian viscoelastic paste was mixed initially through the centrifugal planetary mixer. To ensure effective dispersion and homogenisation of the highly entangled MWCNTs within the polymer matrix, a three-roll milling machine was used to perform the mixing for about 30 mins. The rheology properties of the CNT-PDMS composite pastes, such as the apparent viscosity, was measured by a temperature-controlled rotational rheometer in parallel plate geometry. It was f

关键词： Computer science   Mathematics  

摘要： The advent of the novel Coronavirus Disease 2019 (COVID-19) and itsevolution into a global pandemic has changed the lives of people across theglobe in multiple aspects. In this thesis, we look at one such aspect thathas changed for all of us in the face of the pandemic, grocery shopping. Inparticular, we use computational models to simulate the movement of people in stores and use these models to evaluate the efficacy of a measure thatmost grocery and convenience stores employed at the height of the pandemic,directing aisles. Our experiments simulate how well or badly directed aislesperform in comparison to undirected ones if each agent (shopper) enteringthe store is selfish, and then perform a similar comparison when agents actresponsibly. Our experiments show that selfish agents cause directed aislesto drastically under perform, while if each agent were to act responsibly,directed aisles benefit the agents by allowing them to finish their tasks andleave the store quicker. We then propose several formulations of theoreticaloptimization problems that are constrained to minimize the probability oftransmission of infectious disease. One such model, Staircase-Packing is explored in detail and this problem is proved to be NP-Complete by a reductionfrom the Hamiltonian Path Problem.

关键词： Mathematics   Computer science  

摘要： With the rise of big data analytics, multi-layer neural networks have surfaced as one of the most powerful machine learning methods. Stochastic gradient descent (SGD) is often applied to train Deep Neural Networks (DNNs), and research efforts have been devoted to investigate the convergent dynamics of SGD and minima found by SGD. The influencing factors identified in the literature include learning rate, batch size, Hessian, and gradient covariance, and stochastic differential equations are used to model SGD and establish the relationships among these factors for characterizing minima found by SGD. It has been found that the ratio of batch size to learning rate is a main factor in highlighting the underlying SGD dynamics; however, the influence of other important factors such as the Hessian and gradient covariance is not entirely agreed upon. The factors and relationships in the recent literature are described and numerical findings are presented on the relationships. In particular, it confirms the four-factor and general relationship results obtained in Wang (2019), while the three-factor and associated relationship results found in Jastrzebski et al. (2018) may not hold beyond the considered special case. In addition, the theoretical mathematical properties of multi-layer neural networks are still not fully understood. We seek to extend the mean field results of Mei, Montanari, and Nguyen (2018) from two-layer neural networks with one hidden layer to three-layer neural networks with two hidden layers. Namely, we will show that the SGD dynamics is captured by a set of non-linear partial differential equations, and prove that the distributions of weights in the two hidden layers are independent. We will also detail exploratory work done based on simulation and real-world data.

关键词： Applied Mathematics   Computer science   Molecular physics  

摘要： The Brownian motion of particles can be decomposed into translational and rotational components, characterized by translational and rotational diffusion coefficients respectively. Measuring these two coefficients plays an important role in determining the structure and understanding the dynamic properties of materials, with benefit to research in such areas as molecular biology, material sciences. One of the promising tools for investigation of Brownian motion is the emerging X-ray Photon Correlation Spectroscopy (XPCS), which can capture dynamics of samples comprising large groups of particles in a broad range of time scales and length scales. Methods for estimating translational diffusion coefficients on the basis of the temporal auto-correlation analysis of XPCS images are used widely. However, to the best of our knowledge, there is no XPCS-based algorithm for assessing the rotational diffusion coefficients from such temporal auto-correlation data. In this thesis, we take a different route, and propose exploiting an angular-temporal cross-correlation function whose values are approximated by an estimator based on the collected experimental images. We prove the consistency of this estimator by deriving a tail bound. A numerical algorithm, MTECS, for estimating the rotational diffusion coefficients from the cross-correlation is designed and implemented. We demonstrate the capability of this algorithm by testing it on a range of simulated data.

关键词： Engineering   Artificial intelligence   Computer science  

摘要： Modern networking and computing systems have become very complicated and highly dynamic, which makes them hard to model, predict and control. In this thesis, we aim to study system control problems from a whole new perspective by leveraging emerging Deep Reinforcement Learning (DRL), to develop experience-driven model-free approaches, which enable a network or a device to learn the best way to control itself from its own experience (e.g., runtime statistics data) rather than from accurate mathematical models, just as a human learns a new skill (e.g., driving, swimming, etc). To demonstrate the feasibility and superiority of this experience-driven control design philosophy, we present the design, implementation, and evaluation of multiple DRL-based control frameworks on two fundamental networking problems, Traffic Engineering (TE) and Multi-Path TCP (MPTCP) congestion control, as well as one cutting-edge application, resource co-scheduling for Deep Neural Network (DNN) models on mobile and edge devices with heterogeneous hardware. We first propose DRL-TE, a DRL-based framework that enables experience-driven networking for TE. DRL-TE maximizes a widely-used utility function by jointly learning network environment and its dynamics, and making decisions under the guidance of powerful DNNs. We propose two new techniques, TE-aware exploration and actor-critic-based prioritized experience replay, to optimize the general DRL framework particularly for TE. Furthermore, we propose an Actor-Critic-based Transfer learning framework for TE, ACT-TE, which solves a practical problem in experience-driven networking: when network configurations are changed, how to train a new DRL agent to effectively and quickly adapt to the new environment. In the new network environment, ACT-TE leverages policy distillation to rapidly learn a new control policy from both old knowledge (i.e., distilled from the existing agent) and new experience (i.e., newly collected samples). In addition, we prop

关键词： Robotics   Electrical engineering   Artificial intelligence   Computer science  

关键词： Computer science   Artificial intelligence  

关键词： Algorithms   Optimization   Neural networks   Artificial intelligence   Computer science  

摘要： A trend in computer vision today is to retire older, so-called “classic” methods in favor of ones based on deep neural networks. This has led to tremendous improvements in many areas, but for some problems deep neural solutions may not yet exist or be of practical application. For this and other reasons, classic methods are still widely used in a variety of applications. This paper explores the possibility of using deep neural networks to improve these older methods instead of replace them. In particular, it addresses the issue of parameter selection in these algorithms by using a neural network to predict effective settings on a per-input basis. Specifically, we look at a straightforward and well-understood algorithm with one primary parameter: interactive graph-cut segmentation. This parameter balances region/boundary influences and heavily influences the resulting segmentation. Many approach tuning this parameter by using an ad hoc or empirically selected static setting, while others pre-analyze images to determine effective settings on a per-image basis. Tuning this parameter for each image, or even for each target selection within an image, is highly sensitive to properties of the image and object, suggesting that a network might be able to recognize these properties and predict settings that would improve performance. We employ a lightweight network with minimal layers to avoid adding significant computational overhead with this pre-analysis step. The network predicts the segmentation performance for each of a set of discretely sampled values for this parameter and selects the one with the highest predicted performance. Results demonstrate that this per-image prediction and tuning performs better than a single empirically selected setting

关键词： Design   Analytical chemistry   Atomic physics   Chemistry   Computer science   Energy   Engineering   Physics  

摘要： The safe management of spent nuclear fuel (SNF) is a key challenge facing the nuclear industry today. In the potential event of a premature pinhole cladding failure, fuel pond water can come into contact with highly active spent fuel and remain unnoticed going into storage. In such scenarios, there are risks from reactions including water radiolysis on the fuel surface leading to an accelerated dissolution of the fuel matrix. While this mechanism has been widely studied, there still exist gaps in the understanding of water carryover on dried fuel, precise corrosion mechanisms, and the corrosion rate dependence on lattice modifications in the fuel matrix. Also, the long-term safety implications of such fuel-water interactions are yet to be determined. This thesis addresses the scenario by looking at two pathways of this fuel-water interaction - adsorption and radiolytic dissolution: in three separate strands. In the first strand, wetting of SNF surfaces is considered at high temperatures to evaluate the hygroscopic nature of failed SNF, post drying. This is studied by conducting adsorption isotherm experiments in an experimental rig that was designed and assembled by the author. In the second strand, the radiolytic corrosion mechanism of highly active idealistic SNF surfaces is studied by using epitaxial thin film samples in artificial radiolytic environments. The samples included (U, Ce)O2 in various U:Ce percentage compositions in order to probe the dissolution characteristics of (U, Pu)O2 mixed oxide (MOX) surfaces as a function of induced lattice modifications in the SNF (UO2) matrix. The last strand addresses the scenario of a failed fuel exposed to spent fuel pond water, resulting in wet, waterlogged fuel going into storage. Here static de-ionized (DI) water leaching tests are performed on High Burnup SNF to understand the trends in fuel dissolution as a function of time and surface area of fuel exposed. Further, the fuel surfaces, post leaching, are metallogra

关键词： Computer science   Medical imaging  

摘要： Nuclei segmentation is an important task in medical image analysis. Recently, new deep learning methods have been applied to segmentation resulting in significantly improved performance. However, these applications are constrained by the need for large, balanced datasets. Medical images are often very expensive to annotate and require the valuable time of experts. Data augmentation is often used as an automated way of adding additional instances to address this problem. In this thesis, we propose a method that generates medical images and transfers style between images in addition to random rotation and reflections. Experiments show that our synthetic images improve the accuracy of nuclei segmentation tasks.